An adaptive multi-path computation framework for centrally controlled networks

The Adaptive Dynamic Multi-Path Computation Framework (ADMPCF) is to provide an integrated resource control and management platform with an adequate set of management applications for better routing and resource allocation in centrally controlled or loosely coupled distributed software defined networking (SDN), especially for large network systems. As an open and extensible solution framework, it can provide the necessary infrastructure and integrates data collection and analytics, network performance evaluation, and various optimization algorithms. ADMPCF utilizes a set of complementary algorithms that work together in an adaptive and intelligent fashion that enable global routing and resource allocation optimization. It can also be easily extended to incorporate new algorithms through some open APIs. Such an approach would be able to efficiently and effectively adapt to the rapid changes in network topology, states, and most importantly application traffic, while it is often infeasible for a single optimization algorithm to get satisfactory solution for multiple nonlinear optimization objectives and constraints for a large and centrally controlled network. As it would be costly for centrally controlled global optimization algorithms to calculate good routes dynamically with adequate response time, the proposed ADMPCF framework takes advantage of many hidden patterns of the network information fragments in the combinations of network topology, states, and traffic flows. Therefore, it can lead to a much improved data structure for fast search and match that avoids the expensive re-optimization whenever possible.